> 
> 
> Remark: the current pom:validate goal works to some extent 
> even for poms that extend others. Having a valid 
> a/project.xml and a b/project.xml with only 
> <extend>../a/project.xml</extend> in it, validates b. Making 
> an invalid change to a's pom then makes the validation on b 
> fail as well.

Good feature. I didn't used it.

> 
> However, it doesn't work if I set some minOccurs in the 
> schema file to 1, in that case the validation of b will fail 
> even if the corresponding element is present in a.
> 
> 
> What about the following suggestion: we introduce two 
> schemas, maven-pom-strict and maven-pom-extend. Every 
> stand-alone pom has to comply with pom-strict and every pom 
> that extends another one has to extend at least one strictly 
> valid pom.
> 
> [This doesn't eliminate some exotic setup's where you have a 
> chain of poms that are each invalid by themselves, but when 
> merged together make up a valid one. I don't know if this is 
> a frequent case.]

We should maintain 2 schemas :-(
Or can we generate the maven-pom-extend from maven-pom-strict.

I'm not sure if it's very useful.....

Arnaud

> 
> 
> -Lukas
> 
> 
> Arnaud HERITIER wrote:
> > The only solution is to do what is already done actually...
> > 
> > 1) We can't have a schema more restrictive because each 
> element is optional if you use the POM's inheritence.
> > 2) We can't easily merged project.xml files because there 
> are several 
> > rules : some elements replace the ones in the parent and 
> some others are in addition to them (and it will as 
> complicated as the 3rd proposal).
> > 3) We can't test the rewritten POM because it will not be 
> easy to explain to the user where the problem is ...
> > 
> > The real question we should ask to ourselves it's why we 
> need a schema for the POM ?
> > After thinking about it I'm not sure it's a good idea to 
> validate the 
> > POM with a schema :-( The schema (in my opinion) should be used 
> > exclusively to help users to create the project.xml file 
> (using an xml editor which propose it if they don't already 
> use a tool like mevenide).
> > The validation of the POM should be done by the core itself (with a 
> > validate method in the Project object -even if I don't know what we 
> > can test) and an explicit message should be return if 
> there's a problem. ==> It should be also the model which log 
> a message if a plugin tries to use an element in the POM 
> which is not defined (The test shouldn't be done in the plugin) ....
> > 
> > But I think that all these ideas aren't possible to 
> implement to keep 
> > the compability between maven 10 and maven 1.1 :-( Thus I 
> think we must keep things as more simple as possible.
> > 
> > WDYT ?
> > 
> > Arnaud
> > 
> 
> ---------------------------------------------------------------------
> To unsubscribe, e-mail: [EMAIL PROTECTED] For 
> additional commands, e-mail: [EMAIL PROTECTED]
> 




---------------------------------------------------------------------
To unsubscribe, e-mail: [EMAIL PROTECTED]
For additional commands, e-mail: [EMAIL PROTECTED]

Reply via email to